CN106097043A - The processing method of a kind of credit data and server - Google Patents
The processing method of a kind of credit data and server Download PDFInfo
- Publication number
- CN106097043A CN106097043A CN201610387187.4A CN201610387187A CN106097043A CN 106097043 A CN106097043 A CN 106097043A CN 201610387187 A CN201610387187 A CN 201610387187A CN 106097043 A CN106097043 A CN 106097043A
- Authority
- CN
- China
- Prior art keywords
- instance
- related data
- time related
- terminal
- unit
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 21
- 230000002123 temporal effect Effects 0.000 claims abstract description 81
- 238000000034 method Methods 0.000 claims abstract description 74
- 238000011156 evaluation Methods 0.000 claims abstract description 64
- 230000000875 corresponding effect Effects 0.000 claims description 61
- 230000008569 process Effects 0.000 claims description 44
- 238000012545 processing Methods 0.000 claims description 15
- 238000012937 correction Methods 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 10
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 230000002596 correlated effect Effects 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 230000006399 behavior Effects 0.000 description 49
- 238000012549 training Methods 0.000 description 23
- 238000010606 normalization Methods 0.000 description 18
- 238000013210 evaluation model Methods 0.000 description 12
- 238000010801 machine learning Methods 0.000 description 11
- 230000006870 function Effects 0.000 description 9
- 238000004891 communication Methods 0.000 description 7
- 238000004590 computer program Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 7
- 238000003860 storage Methods 0.000 description 7
- 241001269238 Data Species 0.000 description 6
- 230000006855 networking Effects 0.000 description 6
- 230000001143 conditioned effect Effects 0.000 description 4
- 238000003066 decision tree Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000012216 screening Methods 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 2
- 230000003542 behavioural effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000001413 cellular effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000007935 neutral effect Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000003442 weekly effect Effects 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000012636 effector Substances 0.000 description 1
- 238000013100 final test Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000007261 regionalization Effects 0.000 description 1
- 238000009394 selective breeding Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0609—Buyer or seller confidence or verification
Abstract
The embodiment of the invention discloses a kind of credit data processing method, the method may include that the very first time related data obtained obtained by first instance performs operation in terminal, and this very first time related data is used for characterizing reliable incidence relation between first instance and temporal information；Carrying out latent structure according to very first time related data, generate the first model, this first model is for assessing the credibility of first instance；Obtain active user's behavior of first instance, use record input the first model as terminal to be detected, obtain the credit evaluation result to first instance active user's behavior.The embodiment of the present invention additionally provides a kind of server.
Description
Technical field
The present invention relates to the Internet technology in communications applications field and big data processing technique, particularly relate to a kind of credit
The processing method of data and server.
Background technology
Along with the high speed development of Internet technology, the credit rating of the entity that terminal is corresponding or credit appraisal have had become as use
Family important component part in life application, wherein, above-mentioned entity, for characterizing the object of credit rating to be assessed, such as, is used
Family (individual) or user account number etc..Such as, during service trade working individuality comes into the market circulation personally, other
Consumer selects working individuality when servicing, can with reference to the credit rating that this working is individual determine whether Delegation Server to this from
Industry is individual.
In prior art, server is obtained by the Internet, client or the third-party institution and is used for evaluating individual subscriber (in fact
Body) the original credit data of credit, by this original credit data being analyzed the credit rating of user.Such as, consumer is passed through
The service quality that working is individual is commented on by service for life Vertical Website, or is given based on self assessment standard by mechanism
Marking or the certification credit rating individual to analyze working.
But, when using prior art to carry out the process of credit data, due to for the entity of different application scene,
Original credit data needed for this entity is not quite similar, and therefore, server analyzes credit rating by original credit data
Workload is more complicated, and, the source of the original credit data that this server obtains entity is all based on user behavior
The data of pattern formation, the source causing original credit data is single, affects the letter according to original credit data assessment entity
The effectiveness of expenditure.
Summary of the invention
For solving above-mentioned technical problem, embodiment of the present invention expectation provides processing method and the service of a kind of credit data
Device, it is possible to carried out the unified Analysis of the credit rating of entity by the credit data with same characteristic features of feedback by all kinds of means, real
Show effective assessment of the credit of entity, when using simple Time Correlation Data to carry out credit data process, reduced meanwhile
The workload that data process.
The technical scheme is that and be achieved in that:
A kind of credit data processing method that the embodiment of the present invention provides, including:
Obtaining the very first time related data obtained by first instance performs operation in terminal, the described very first time is correlated with
Data are used for characterizing reliable incidence relation between first instance and temporal information；
Carrying out latent structure according to described very first time related data, generate the first model, described first model is used for commenting
Estimate the credibility of first instance；
Obtain active user's behavior of described first instance, use record input described the as terminal to be detected
One model, obtains the credit evaluation result to first instance active user's behavior.
In such scheme, described acquisition first instance performs the very first time dependency number obtained by operation in terminal
According to, including:
Obtain the original time related data obtained by described first instance performs operation in described terminal；
Described original time related data is carried out pretreatment according to preset strategy, with from described original time related data
In filter out invalid non-temporal information, obtain comprising the very first time related data of effective temporal information.
In such scheme, described described original time related data is carried out pretreatment according to preset strategy, including:
Be removed improper described original time related data processing, to same terminal in time of certain length
The described original time related data that in section, number of repetition is too much carries out duplicate removal or to abnormal described original time related data
It is purged processing, to filter out invalid non-temporal information from described original time related data, when obtaining comprising effective
Between the described very first time related data of information.
In such scheme, described acquisition first instance performs the very first time dependency number obtained by operation in terminal
According to, including:
Described terminal is specified to inform the allowed band of user's data to be collected；
After described terminal obtains the license of described user, obtain described first instance and perform to operate in described terminal
That arrive and in described allowed band described very first time related data.
In such scheme, described carry out latent structure according to described very first time related data, including:
The first use temporal information is extracted in described very first time related data；
Use temporal information according to described first, add up described in the described allowed band in presetting each time period
The frequent degree of use of first instance and use distributed intelligence；
Carry out according to described very first time related data, the frequent degree of use of described first instance and use distributed intelligence
Latent structure.
In such scheme, described carry out latent structure according to described very first time related data, generate the first model it
After, described method also includes:
Feature based on existing subscriber's behavior and corresponding existing credit result, be modified described first model,
To the second model；
Accordingly, active user's behavior of the described first instance of described acquisition, use note as terminal to be detected
Record inputs described first model, obtains the credit evaluation result to first instance active user's behavior, including:
Obtain described active user's behavior of described first instance, record defeated as described terminal use to be detected
Enter described second model, obtain the described credit evaluation result to described first instance active user's behavior.
In such scheme, described in obtain the credit evaluation result to first instance active user's behavior after, described side
Method also includes:
According to pre-set criteria, determine the second instance associated with described first instance；
According to described first instance and the pre-set priority of described second instance, determine weights and the institute of described first instance
State the weights of second instance；
The credit evaluation result of described first instance, the weights of described first instance, the credit of described second instance are commented
The weights input estimating result and described second instance is tied to default 3rd model, the correction credit evaluation obtaining described first instance
Really.
Embodiments provide a kind of server to include:
Acquiring unit, for obtaining the very first time related data obtained by first instance performs operation in terminal, institute
State very first time related data for characterizing reliable incidence relation between first instance and temporal information；
Structural unit, carries out latent structure for the described very first time related data obtained according to described acquiring unit,
Described signal generating unit generates the first model, and described first model is for assessing the credibility of first instance；
Described acquiring unit, is additionally operable to obtain active user's behavior of described first instance；
Output unit, for active user's behavior of described first instance of being obtained by described acquiring unit as to be detected
Terminal use record input described signal generating unit generate described first model, obtain first instance active user's behavior
Credit evaluation result.
In above-mentioned server, described server also includes: pretreatment unit；
Described acquiring unit, be additionally operable to obtain described first instance performs obtained by operation in described terminal original time
Between related data；
Described pretreatment unit, the described original time related data being used for obtaining described acquiring unit is according to default plan
Slightly carry out pretreatment；
Described acquiring unit, specifically for filtering out through described pretreatment unit from described original time related data
Invalid non-temporal information, obtains comprising the very first time related data of effective temporal information.
In above-mentioned server, described pretreatment unit, specifically for the institute obtaining improper described acquiring unit
State original time related data to be removed processing, to the too much institute of same terminal number of repetition within the time period of certain length
State original time related data carry out duplicate removal or be purged processing, with from institute to abnormal described original time related data
State and original time related data filters out invalid non-temporal information, obtain comprising the described very first time of effective temporal information
Related data.
In above-mentioned server, described server also includes: designating unit；
Designating unit, for specifying described terminal to inform the allowed band of user's data to be collected；
Described acquiring unit, specifically for when after the license of the described terminal described user of acquisition, obtaining described first instance
Described terminal performs that operation obtains and in the described allowed band that described designating unit is specified the described very first time
Related data.
In above-mentioned server, described server also includes: extraction unit, statistic unit；
Described extraction unit, makes for extracting first in the described very first time related data that described acquiring unit obtains
Use temporal information；
Described statistic unit, described first for extracting according to described extraction unit uses temporal information, and statistics is in advance
If the frequent degree of use of the described first instance in the described allowed band that the described designating unit in each time period is specified
With use distributed intelligence；
Described structural unit, specifically for the described very first time related data obtained according to described acquiring unit, described
The frequent degree of use and the use distributed intelligence of the described first instance of statistic unit statistics carry out latent structure.
In above-mentioned server, described server also includes: amending unit；
Described amending unit, carries out latent structure, institute for described structural unit according to described very first time related data
Stating after signal generating unit generates the first model, feature based on existing subscriber's behavior and corresponding existing credit result, to described
Described first model that signal generating unit generates is modified, and obtains the second model；
Described acquiring unit, also particularly useful for the described active user's behavior obtaining described first instance；
Described output unit, active user's behavior of the described first instance specifically for being obtained by described acquiring unit is made
Use record to input described second model that described signal generating unit generates for described terminal to be detected, obtain first instance is worked as
The credit evaluation result of front user behavior.
In above-mentioned server, described server also comprises determining that unit；
Described determine unit, obtain the credit evaluation result to first instance active user's behavior for described output unit
Afterwards, according to pre-set criteria, the second instance associated with described first instance is determined；And according to described first instance and described
The pre-set priority of second instance, determines weights and the weights of described second instance of described first instance；
Described output unit, is additionally operable to the credit evaluation result of described first instance, the described institute determining that unit determines
State the weights of first instance, the credit evaluation result of described second instance and described determine described second instance that unit determines
Weights input, to presetting the 3rd model, obtains the correction credit evaluation result of described first instance.
Embodiments provide processing method and the server of a kind of credit data, by obtaining first instance at end
Performing the very first time related data obtained by operation on end, this very first time related data is used for characterizing first instance and time
Reliable incidence relation between information；Carry out latent structure according to very first time related data, generate the first model, this first mould
Type is for assessing the credibility of first instance；Obtain active user's behavior of first instance, as terminal to be detected
Use record input the first model, obtain the credit evaluation result to first instance active user's behavior.Use above-mentioned technology real
Existing scheme, the multiple entity Time Correlation Data in server by utilizing terminal, build one's credit in conjunction with feedback information by all kinds of means
Evaluation model, it is possible to effectively obtain reflecting the index of entity confidence level, i.e. credit evaluation result, it is achieved related entities is believed
Effective assessment；And use terminal use when present in relatively simple temporal information as key data source, in advance
Processing procedure and latent structure process are the most simple, it is not necessary to use various complexity coding, cluster, screening means are to feature
Carry out structure and the process of complexity, thus greatly reduce the workload that data process so that model and system are the most available.
Accompanying drawing explanation
Fig. 1 is the schematic diagram carrying out the mutual various hardware entities of information in the embodiment of the present invention；
The block schematic illustration one of a kind of credit data processing method that Fig. 2 provides for the embodiment of the present invention；
The block schematic illustration two of a kind of credit data processing method that Fig. 3 provides for the embodiment of the present invention；
The training process schematic of a kind of model that Fig. 4 provides for the embodiment of the present invention；
The block schematic illustration three of a kind of credit data processing method that Fig. 5 provides for the embodiment of the present invention；
The structural representation one of a kind of server that Fig. 6 provides for the embodiment of the present invention；
The structural representation two of a kind of server that Fig. 7 provides for the embodiment of the present invention；
The structural representation three of a kind of server that Fig. 8 provides for the embodiment of the present invention；
The structural representation four of a kind of server that Fig. 9 provides for the embodiment of the present invention；
The structural representation five of a kind of server that Figure 10 provides for the embodiment of the present invention；
The structural representation six of a kind of server that Figure 11 provides for the embodiment of the present invention；
The structural representation seven of a kind of server that Figure 12 provides for the embodiment of the present invention.
Detailed description of the invention
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Describe wholely.
In order to preferably introduce and understand various embodiments of the present invention, being described below can in various embodiments of the present invention
Some specialized vocabularies that can relate to, specifically include:
Terminal: refer to mobile electronic device, also referred to as running gear (English: Mobile device), flow device, hands
Hold device (handheld device), wearable device etc., be a kind of calculating equipment based on embedded chip, generally have one
Individual little display screens, touch-control inputs, or small-sized keyboard.
Entity: terminal itself, is also contained in terminal the account number used, and also includes the actual effector of this terminal and has
Person.
Machine learning: rely on theory of probability, statistics, neural propagation scheduling theory, enable a computer to the study of simulating human
Behavior, to obtain new knowledge or technical ability, reorganizes existing knowledge structure and is allowed to constantly improve the performance of self.
Model training: the sample of artificial selection is inputed to machine learning system, by constantly adjusting model parameter, makes
Final cast to the rate of accuracy reached of specimen discerning to optimum.
Credit evaluation result: according to training pattern and the credit value of the entity of history reality, new entity is estimated
The credit result of meter.
As it is shown in figure 1, for the embodiment of the present invention carries out showing of various hardware entities in the system architecture that information is mutual
Being intended to, Fig. 1 includes: one or more servers 41～4n, terminal unit 21-25 and network 31, and network 31 includes route
Device, gateway etc. network entity, does not embody in Fig. 1.Terminal unit 21-25 is by cable network or wireless network and service
It is mutual that device 41～4n carries out service product information, in order to obtains from terminal 21-25 and produces Time Correlation Data by user behavior
And transmit to server 41～4n.The type of terminal unit is as it is shown in figure 1, include mobile phone (terminal 23), panel computer or PDA
The type such as (terminal 25), desktop computer (terminal 22), PC (terminal 24), all-in-one (terminal 21).Wherein, terminal unit is installed
Have the applied function module needed for various user, such as possess amusement function application (such as Video Applications, audio frequency plays application,
Game application, ocr software), the application and for example possessing service function (such as digital map navigation application, purchases by group application, camera application
Deng), furthermore the systemic-functions such as application are such as set.
Based on the hardware entities shown in above-mentioned Fig. 1, user uses terminal by using the application in terminal or terminal to produce
Use data accordingly, these concrete use data are very first time related data, and server is obtained in this terminal by terminal
Very first time related data to carry out effective pretreatment regular, and combine associated credit record and the knowledge of first instance,
And passing credit record, it is proposed that the credit of a kind of first instance to being associated with terminal is evaluated, and obtains credit
The method of assessment result.The embodiment of the present invention provide credit data processing method be based on server based on first instance time
Between related data, by Credit Evaluation Model, the i.e. first model, evaluate the process of credit evaluation result.With to be evaluated
As a example by application is for bank client application, server is got when user uses the first account number of bank client by terminal
Use the use data that temporal information and associated credit card thereof are refunded, and set up the first mould of temporal characteristics and credit appraisal value
Type, the credit evaluation result obtaining the credit card repayment to this first account number is that credit is good.
The example of above-mentioned Fig. 1 simply realizes a system architecture example of the embodiment of the present invention, and the embodiment of the present invention is not
It is limited to the system structure described in above-mentioned Fig. 1, based on this system architecture, proposes each embodiment of the present invention.
Embodiment one
Embodiments provide a kind of credit data processing method, as in figure 2 it is shown, the method may include that
S101, acquisition first instance perform the very first time related data obtained by operation, this very first time in terminal
Related data is used for characterizing reliable incidence relation between first instance and temporal information.
It should be noted that a kind of credit data processing method that the embodiment of the present invention provides is that server side is by from end
The use time data relevant to entity that end obtains, carries out the process of credit appraisal to entity itself.
Particularly, the embodiment of the present invention is the mistake of the credit appraisal launched around terminal based on very first time related data
Journey.
It should be noted that the credit data that very first time related data is first instance in the embodiment of the present invention.
Optionally, the first instance in the embodiment of the present invention can be hold terminal person, use terminal person, terminal itself or
Each application in person's terminal, such as, social networking application, music application, video display application etc..
Concrete, the process that the very first time of the server acquisition first instance in the embodiment of the present invention first closes data is permissible
For: server designated terminal informs the allowed band of user's data to be collected；After terminal obtains the license of user, this server
Obtain first instance in terminal, perform what operation obtained, and the above-mentioned very first time related data in allowed band.The most just
Being to say, in embodiments of the present invention, when server to carry out the credit appraisal to first instance, server can go to obtain by designated terminal
Take the allowed band of the very first time related data of this first instance, after this terminal gets the allowed band of user's license, just
The very first time related data in allowed band can be obtained.
It should be noted that first instance in the embodiment of the present invention can use in corresponding multiple application scenarios, on
State allowed band and just can characterize the very first time related data that user allows which application scenarios of terminal acquisition first instance,
I.e. server can get the very first time related data of which application scenarios corresponding to first instance, and therefore, server obtains
The credit data of the first instance taken can be the very first time related data in multiple allowed band, and it is real that server obtains first
The Time Correlation Data which application scenarios of body is corresponding is determined by allowed band.
Exemplary, server can obtain M the very first time dependency number corresponding with M allowed band of first instance
According to, wherein, M is more than or equal to 1.Assume user account 123, be applied in the wechat in social networking application, be also applied to pay application
In Alipay in.When first instance is at the user account 123 of the enterprising enforcement of terminal, user allow terminal to wechat and
Alipay (allowed band) carries out data access collection, and then this server gets the letter wechat uses scene from terminal
With data and the very first time related data in Alipay uses scene, i.e. server gets 2 allowed band correspondences
Very first time related data.
Optionally, the very first time related data in the embodiment of the present invention may include that the first use temporal information and
One service data.First use time for be characterized in terminal carry out relevant to first instance dissimilar operate time time
Between, the first service data for characterize carry out first instance credit rating estimate the content that required different allowed bands are corresponding
Data.
Optionally, the terminal in the embodiment of the present invention obtains corresponding each first the making of each allowed band of first instance
It is at least one in scenario described below by temporal information and each first service data:
(1), user uses time point that account number (first instance) carries out logging in terminal, (first uses to log in duration
Temporal information), and the title of relevant account number or describe information (the first service data)；
(2), user's browser application (first instance) in terminal etc. use network to connect and the time of network service
Segment information (first uses temporal information), and use the link information of network connection, the agreement using network to connect, use net
The content of network service and service provider information (the first service data)；
(3), user uses terminal such as global positioning system (GPS, Global Positioning System) application (the
One entity) carry out the positional information (the of the time point of satellite fix, duration information (first use temporal information) and location
One service data) etc.；
(4), user by terminal by intelligent use thereon (first instance) control other electronic equipments time point,
Duration (first uses temporal information), and the type information (the first service data) of other electronic equipments；
(5), user uses time point that the communications applications (first instance) in terminal carries out exchanging with other people, when continuing
The relevant information (the first service data) such as long (first uses temporal information) and use data, such as chat record, exchange way
Include but not limited to mobile cellular telephone, note, the networking telephone, chat software, social networks, video calling etc.；
(6), user use terminal pass through to pay application (first instance) carry out paying or the time point of certification (first uses
Temporal information), and pay or the relevant information (the first service data) of certification；
(7), user's media application (first instance) in terminal carries out the culture and recreation movable time period (first makes
By temporal information), and the content description information (the first service data) of activity.Cultural activity mainly includes playing in multimedia
Holding, receive and use Streaming Media, recreation mainly includes mobile electron game, social gaming, online game etc.；
(8), the model of terminal (first instance), brand, device name, unique recognition coding IMEI/SN/MEID (first
Service data), and information correlation time (first uses temporal information) that battery uses；
(9), user adjust the relevant setting of terminal (first instance), arrange including system, network connect arrange, interface is handed over
The time point information (first uses temporal information) that arranges mutually and relevant parameter (the first service data) is set.
It is to say, terminal can be for the concrete practical situation of first instance, in the above-mentioned allowed band of acquisition request
First uses the very first time related data such as temporal information and the first service data, and this very first time dependency number that will get
According to uploading onto the server.
Concrete, the very first time related data collected can be uploaded to service by the way of cloud storage by terminal
Device.
It should be noted that above-mentioned each application scenarios can a corresponding very first time related data, the present invention is real
Execute the very first time related data in example and refer to difference correspondence in the first instance all application scenarios in the range of allowing you
The summation of very first time related data.
S102, carrying out latent structure according to very first time related data, generate the first model, this first model is used for assessing
The credibility of first instance.
After server obtains very first time related data, this server just can build according to very first time related data
The first service data in very first time related data and the first fisrt feature using time correlation.
Obtain it should be noted that the very first time related data in the embodiment of the present invention can be server from terminal
Original time related data, it is also possible to be these original time data after pretreatment, that disposes after invalid information is effective
Data.Concrete, very first time related data be server by original time related data through the side of pretreated data
Formula will be described in detail in subsequent embodiment.
It should be noted that owing to the allowed band of the very first time related data of first instance can be multiple applied field
Scape, therefore, the Time Correlation Data that each application scenarios in very first time related data is corresponding in embodiments of the present invention
One feature of corresponding structure, say, that be the feature of various dimensions according to the feature of very first time related data structure.
Exemplary, it is assumed that the allowed band of first instance is M application scenarios.Very first time related data just includes M
Individual first uses temporal information and M the first service data, and this server just can use temporal information and M according to M first
Individual first service data, builds each first service data based on first instance first and uses the fisrt feature of time correlation,
M fisrt feature corresponding to above-mentioned M service data just constitutes fisrt feature storehouse, namely described in the embodiment of the present invention
Feature, this completes server and carries out the process of latent structure according to very first time related data, and wherein, M is more than or equal to
1。
Concrete, in a kind of credit data processing method provided in the embodiment of the present invention, server is according to the very first time
Related data carries out the concrete grammar of latent structure and may include that extracting for the first use time in very first time related data believes
Breath；Server uses temporal information according to first, the first instance in statistics allowed band in presetting each time period
Use frequent degree and use distributed intelligence；Server is according to the frequent degree of use of very first time related data, first instance
Latent structure is carried out with using distributed intelligence.
Optionally, the frequent degree that uses of the first instance in the embodiment of the present invention referred in H the default time period
Each application scenarios in the access times that used of this first instance, use distributed intelligence to refer in each application scenarios the
The first quantity that the use in above-mentioned default H the time period of two service datas is non-zero.
Optionally, default H the time period in the embodiment of the present invention is can arrange, and concrete can be set voluntarily by user
Put, it is also possible to for the unified setting of acquiescence, it is also possible to dynamically regulate according to different practical situations, when concrete default H
Between section divide and the determination embodiment of the present invention of concrete numerical value is not restricted.Such as, during default H in the embodiment of the present invention
Between section can be to be adjusted from big to small according to natural time unit, usually the moon, week, day, hour and minute order
It is adjusted.
Particularly, the number of H is at least two, and this is due to the accuracy of the credit appraisal in order to ensure first instance,
Data statistics process to first instance as much as possible wants detailed and complete, and the embodiment of the present invention is not intended to the numerical value of H, can fit
Answering property adjusts.
It should be noted that fisrt feature storehouse described above is exactly the description of first instance and the various dimensions of time correlation
The data set of the user behavior feature within default H the time period.The composition in fisrt feature storehouse is the M with first instance first
Use temporal information relevant with M the first service data, therefore, the integrity of the very first time related data of this first instance
The highest, then the fisrt feature storehouse of its correspondence is the highest for the accuracy rate of follow-up determination credit evaluation result.
As a example by the allowed band of the first instance in the embodiment of the present invention is for M application scenarios below, real in the present invention
Executing in a kind of credit data processing method that example provides, server is when building the fisrt feature storehouse relevant to first instance, sharp
Temporal information is used, to individual first service data of M of each time period in default H the time period (based on user with M first
The data that behavior uses) carry out data statistics, at least constitute M × H dimension with time correlation fisrt feature storehouse.
Concrete, server can use temporal information according to M first, and M the first service data of statistics is at default H
M the first access times corresponding in each time period in time period and M distributed intelligence；According to M the first service data
Pre-set priority, determine the M corresponding with M the first service data the first weights；M in each time period first is made
It is weighted mapping with number of times and M the first weights, obtains M and use weighted value；Weighted value and M distribution is used according to M
Information, constructs the characteristic vector of each time period of first instance to be asked；According to each time period in default H the time period
Characteristic vector, construct the fisrt feature storehouse that first instance to be asked is corresponding.
It should be noted that the pre-set priority of M the first service data in the embodiment of the present invention can be for table
Levy the importance information of the second service data.For example, it is assumed that the very first time related data of first instance includes 3 first
Use temporal information and 3 the first service datas, it is assumed that these 3 service datas include: (1), (8) in above-mentioned application scenarios
(9), then run higher than first of the application correspondence in terminal due to the importance of the first service data of terminal correspondence itself
Data, and in the various application in terminal, the importance arranging the first service data corresponding to application of terminal is higher than other merits
Property can apply the first corresponding service data, and (1) is the first service data that other functions application in terminal is corresponding, (8)
For the first service data that the terminal in terminal itself is corresponding, (9) are first service data arranging application correspondence of terminal, because of
This, pre-set priority is: the priority of the first service data in (8), more than the priority of the first service data in (9),
Priority more than the first service data in (1).
Further, server after the fisrt feature storehouse constructing time correlation, for the facility unifiedly calculated and
Conversion, the fisrt feature (above-mentioned characteristic vector) in the above-mentioned fisrt feature storehouse constructed to be normalized by this server
Process.
It is to say, server is according to the characteristic vector of each time period preset in H time period, construct first real
The fisrt feature storehouse that body is corresponding, may include that server according to preset rules, to each time period in default H the time period
Characteristic vector be normalized；The characteristic vector of each time period after normalization is configured to first instance corresponding
Fisrt feature storehouse.
Concrete, in embodiments of the present invention, fisrt feature storehouse is normalized as using normalization by server
The factor is multiplied with the feature in above-mentioned fisrt feature storehouse, it is achieved to the normalized in fisrt feature storehouse.
It should be noted that the determination of the normalization factor in the embodiment of the present invention can have multiple, the embodiment of the present invention
It is not restricted.
Exemplary, the normalization factor in the embodiment of the present invention can be determine according to unit of time；Can also be
Determine according to information type；Can also be determine according to the combination of unit of time and information type.Specific as follows:
(1), it is normalized according to unit of time.Common, server is according to each natural sky, according to weekly or press
According to monthly carrying out the ASSOCIATE STATISTICS (average, median, maximum, summation, percentile etc.) of fisrt feature, then use statistics
The inverse of the maximum of one or more in amount is as normalization factor.Wherein, the result after 0 value normalization is remained as
0。
(2), it is normalized according to information type.The time that comprises that i.e. server based on data is collected listed by step believes
The very first time related data ((8) except above-mentioned) corresponding to each application scenarios of breath, calculates normalization factor respectively.
(3), integrate (1), the method for (2), server by asking larger or smaller in both normalization factor one,
Fisrt feature in fisrt feature storehouse is normalized.
It should be noted that the normalization factor determined by said method, server is by the normalizing in fisrt feature storehouse
Fisrt feature storehouse after change carries out follow-up credit appraisal flow process for the list entries of the first model.
It should be noted that after server carries out latent structure according to very first time related data, this server can be given birth to
Become fisrt feature and for characterizing first instance credibility, there is the first model of corresponding relation, i.e. credit prediction model.
Optionally, structure or the method for generation of the first model in the embodiment of the present invention can be by common machine learning
Sorting technique carry out, such as, support vector machine, logistic regression, decision tree, iteration decision tree (GBDT, Gradient
Boosting Decision Tree) or neutral net.By by the second operation number in first instance in the embodiment of the present invention
According to as target variable, the feature that input has constructed is trained, and adjusts parameter, obtains the first instance that can effectively predict
The optimal models that discreditable behavior probability occurs of the most credible or following a period of time.
S103, active user's behavior of acquisition first instance, use record input first as terminal to be detected
Model, obtains the credit evaluation result to first instance active user's behavior.
After server completes the generation of latent structure and the first model, owing to server has been set up first instance
In the corresponding relation (the i.e. first model) of fisrt feature and credibility, therefore, this server just can acquisition first
Active user's behavior of entity inputs the first model, obtains the credit evaluation result to first instance active user's behavior, thus
Realize the credit appraisal of first instance.
It should be noted that for a terminal, the credit that carries out of the first instance in terminal and other entities is commented
The process of valency is consistent, and other entities in terminal can also obtain its independent credit evaluation result by the first model.
It is understood that in embodiments of the present invention, the behavioral data of the various complexity of existing use, the present invention are compared
Use terminal use when present in relatively simple temporal information as key data source, preprocessing process and latent structure mistake
Journey is the most simple, it is not necessary to use various complexity coding, cluster, screening means feature is carried out complexity structure and place
Reason, greatly reduces the workload that data process so that Credit Evaluation Model is the most available.
Embodiment two
A kind of credit data processing method that the embodiment of the present invention provides, as it is shown on figure 3, the method may include that
S201, acquisition first instance perform the original time related data obtained by operation, this original time in terminal
Related data is used for characterizing reliable incidence relation between first instance and temporal information.
It should be noted that a kind of credit data processing method that the embodiment of the present invention provides is that server side is by from end
The use time data relevant to entity that end obtains, carries out the process of credit appraisal to entity itself.
Particularly, the embodiment of the present invention is the process of the credit appraisal launched around terminal based on Time Correlation Data.
It should be noted that the credit data that original time related data is first instance in the embodiment of the present invention.
Optionally, the first instance in the embodiment of the present invention can be hold terminal person, use terminal person, terminal itself or
Each application in person's terminal, such as, social networking application, music application, video display application etc..
Concrete, the process that the very first time of the server acquisition first instance in the embodiment of the present invention first closes data is permissible
For: server designated terminal informs the allowed band of user's data to be collected；After terminal obtains the license of user, this server
Obtain first instance in terminal, perform what operation obtained, and the above-mentioned original time related data in allowed band.The most just
Being to say, in embodiments of the present invention, when server to carry out the credit appraisal to first instance, server can go to obtain by designated terminal
Take the allowed band of the original time related data of this first instance, after this terminal gets the allowed band of user's license, just
The original time related data in allowed band can be obtained.
It should be noted that first instance in the embodiment of the present invention can use in corresponding multiple application scenarios, on
State allowed band and just can characterize the original time related data that user allows which application scenarios of terminal acquisition first instance,
I.e. server can get the original time related data of which application scenarios corresponding to first instance, and therefore, server obtains
The credit data of the first instance taken can be the original time related data in multiple allowed band, and it is real that server obtains first
The Time Correlation Data which application scenarios of body is corresponding is determined by allowed band.
Exemplary, server can obtain M the original time dependency number corresponding with M allowed band of first instance
According to, wherein, M is more than or equal to 1.Assume user account 123, be applied in the wechat in social networking application, be also applied to pay application
In Alipay in.When first instance is at the user account 123 of the enterprising enforcement of terminal, user allow terminal to wechat and
Alipay (allowed band) carries out data access collection, and then this server gets the letter wechat uses scene from terminal
With data and the original time related data in Alipay uses scene, i.e. server gets 2 allowed band correspondences
Original time related data.
Optionally, the original time related data in the embodiment of the present invention may include that the first use temporal information and
One service data.First use time for be characterized in terminal carry out relevant to first instance dissimilar operate time time
Between, the first service data for characterize carry out first instance credit rating estimate the content that required different allowed bands are corresponding
Data.
Optionally, the terminal in the embodiment of the present invention obtains corresponding each first the making of each allowed band of first instance
It is at least one in scenario described below by temporal information and each first service data:
(1), user uses time point that account number (first instance) carries out logging in terminal, (first uses to log in duration
Temporal information), and the title of relevant account number or describe information (the first service data)；
(2), user's browser application (first instance) in terminal etc. use network to connect and the time of network service
Segment information (first uses temporal information), and use the link information of network connection, the agreement using network to connect, use net
The content of network service and service provider information (the first service data)；
(3), user uses terminal such as GPS application (first instance) to carry out the time point of satellite fix, duration information
(first uses temporal information) and the positional information (the first service data) etc. of location；
(4), user by terminal by intelligent use thereon (first instance) control other electronic equipments time point,
Duration (first uses temporal information), and the type information (the first service data) of other electronic equipments；
(5), user uses time point that the communications applications (first instance) in terminal carries out exchanging with other people, when continuing
The relevant information (the first service data) such as long (first uses temporal information) and use data, such as chat record, exchange way
Include but not limited to mobile cellular telephone, note, the networking telephone, chat software, social networks, video calling etc.；
(6), user use terminal pass through to pay application (first instance) carry out paying or the time point of certification (first uses
Temporal information), and pay or the relevant information (the first service data) of certification；
(7), user's media application (first instance) in terminal carries out the culture and recreation movable time period (first makes
By temporal information), and the content description information (the first service data) of activity.Cultural activity mainly includes playing in multimedia
Holding, receive and use Streaming Media, recreation mainly includes mobile electron game, social gaming, online game etc.；
(8), the model of terminal (first instance), brand, device name, unique recognition coding IMEI/SN/MEID (first
Service data), and information correlation time (first uses temporal information) that battery uses；
(9), user adjust the relevant setting of terminal (first instance), arrange including system, network connect arrange, interface is handed over
The time point information (first uses temporal information) that arranges mutually and relevant parameter (the first service data) is set.
It is to say, terminal can be for the concrete practical situation of first instance, in the above-mentioned allowed band of acquisition request
First uses the original time related data such as temporal information and the first service data, and this original time dependency number that will get
According to uploading onto the server.
Concrete, the original time related data collected can be uploaded to service by the way of cloud storage by terminal
Device.
It should be noted that above-mentioned each application scenarios can a corresponding original time related data, the present invention is real
Execute the original time related data in example and refer to difference correspondence in the first instance all application scenarios in the range of allowing you
The summation of original time related data.
S202, original time related data is carried out pretreatment according to preset strategy, with from original time related data
Filter out invalid non-temporal information, obtain comprising the very first time related data of effective temporal information.
After server obtains the original time related data obtained by first instance performs operation in terminal, due to this
When the original time related data that server in inventive embodiments obtains there may be a certain application scenarios corresponding original
Between the situations such as the existence disappearance of related data, invalid or certain part data duplicate occur, such original time is correlated with
Data to server during the follow-up credit appraisal carrying out first instance meaningless, and increase the work of calculating
Amount.Therefore, server is before carrying out the credit appraisal of first instance, and this server will be to original time related data according in advance
If strategy carries out pretreatment, to filter out invalid non-temporal information from this original time related data, obtain comprising effectively
The very first time related data of temporal information.
Concrete, server is after getting the original time related data that terminal is uploaded, whole to get
Original time related data carries out pretreatment, i.e. server and is filtered out through data cleansing and data by original time related data
Effective very first time related data.Wherein, the standard of data screening is carried out according to the integrated degree of information, for above-mentioned several
The credit data that there is original time related data disappearance in application scenarios should be removed；Data cleansing includes time improper
Between original time related data removal, to the too much original time of number of repetition in the time period of same terminal certain length
Related data carries out duplicate removal, original time related data removing etc. to exceptional value, server by above-mentioned pretreatment original time
Between effective very first time related data is remained as far as possible after related data, then, server can be according to first
Time Correlation Data proceeds the process of the credit appraisal of following first instance.
It is understood that owing to server in embodiments of the present invention carries out the very first time dependency number of credit appraisal
According to being the effective time related data through pretreatment, therefore, the follow-up process number carrying out first instance credit appraisal is being ensured
According to accuracy on the premise of, the workload that the data that decrease process, improve treatment effeciency.
S203, carrying out latent structure according to very first time related data, generate the first model, this first model is used for assessing
The credibility of first instance.
The existing credit result of S204, feature based on existing subscriber's behavior and correspondence, is modified the first model,
To the second model.
S205, active user's behavior of acquisition first instance, use record input second as terminal to be detected
Model, obtains the credit evaluation result to first instance active user's behavior.
After server carries out pretreatment to very first time related data according to preset strategy, owing to this server is the most clear
Remove repetition and invalid original time related data has obtained effectively usable very first time related data, therefore, these clothes
Business device just can use with first according to the first service data that very first time related data builds in very first time related data
The fisrt feature of time correlation.
It should be noted that owing to the allowed band of the very first time related data of first instance can be multiple applied field
Scape, therefore, the Time Correlation Data that each application scenarios in very first time related data is corresponding in embodiments of the present invention
One feature of corresponding structure, say, that be the feature of various dimensions according to the feature of very first time related data structure.
Exemplary, it is assumed that the allowed band of first instance is M application scenarios.Very first time related data just includes M
Individual first uses temporal information and M the first service data, and this server just can use temporal information and M according to M first
Individual first service data, builds each first service data based on first instance first and uses the fisrt feature of time correlation,
M fisrt feature corresponding to above-mentioned M service data just constitutes fisrt feature storehouse, namely described in the embodiment of the present invention
Feature, this completes server and carries out the process of latent structure according to very first time related data, and wherein, M is more than or equal to
1。
Concrete, in a kind of credit data processing method provided in the embodiment of the present invention, server is according to the very first time
Related data carries out the concrete grammar of latent structure and may include that extracting for the first use time in very first time related data believes
Breath；Server uses temporal information according to first, the first instance in statistics allowed band in presetting each time period
Use frequent degree and use distributed intelligence；Server is according to the frequent degree of use of very first time related data, first instance
Latent structure is carried out with using distributed intelligence.
Optionally, the frequent degree that uses of the first instance in the embodiment of the present invention referred in H the default time period
Each application scenarios in the access times that used of this first instance, use distributed intelligence to refer in each application scenarios the
The first quantity that the use in above-mentioned default H the time period of two service datas is non-zero.
Optionally, default H the time period in the embodiment of the present invention is can arrange, and concrete can be set voluntarily by user
Put, it is also possible to for the unified setting of acquiescence, it is also possible to dynamically regulate according to different practical situations, when concrete default H
Between section divide and the determination embodiment of the present invention of concrete numerical value is not restricted.Such as, during default H in the embodiment of the present invention
Between section can be to be adjusted from big to small according to natural time unit, usually the moon, week, day, hour and minute order
It is adjusted.
Particularly, the number of H is at least two, and this is due to the accuracy of the credit appraisal in order to ensure first instance,
Data statistics process to first instance as much as possible wants detailed and complete, and the embodiment of the present invention is not intended to the numerical value of H, can fit
Answering property adjusts.
It should be noted that fisrt feature storehouse described above is exactly the description of first instance and the various dimensions of time correlation
The data set of the user behavior feature within default H the time period.The composition in fisrt feature storehouse is the M with first instance first
Use temporal information relevant with M the first service data, therefore, the integrity of the very first time related data of this first instance
The highest, then the fisrt feature storehouse of its correspondence is the highest for the accuracy rate of follow-up determination credit evaluation result.
As a example by the allowed band of the first instance in the embodiment of the present invention is for M application scenarios below, real in the present invention
Executing in a kind of credit data processing method that example provides, server is when building the fisrt feature storehouse relevant to first instance, sharp
Temporal information is used, to individual first service data of M of each time period in default H the time period (based on user with M first
The data that behavior uses) carry out data statistics, at least constitute M × H dimension with time correlation fisrt feature storehouse.
Concrete, server can use temporal information according to M first, and M the first service data of statistics is at default H
M the first access times corresponding in each time period in time period and M distributed intelligence；According to M the first service data
Pre-set priority, determine the M corresponding with M the first service data the first weights；M in each time period first is made
It is weighted mapping with number of times and M the first weights, obtains M and use weighted value；Weighted value and M distribution is used according to M
Information, constructs the characteristic vector of each time period of first instance to be asked；According to each time period in default H the time period
Characteristic vector, construct the fisrt feature storehouse that first instance to be asked is corresponding.
It should be noted that the pre-set priority of M the first service data in the embodiment of the present invention can be for table
Levy the importance information of the second service data.For example, it is assumed that the very first time related data of first instance includes 3 first
Use temporal information and 3 the first service datas, it is assumed that these 3 service datas include: (1), (8) in above-mentioned application scenarios
(9), then run higher than first of the application correspondence in terminal due to the importance of the first service data of terminal correspondence itself
Data, and in the various application in terminal, the importance arranging the first service data corresponding to application of terminal is higher than other merits
Property can apply the first corresponding service data, and (1) is the first service data that other functions application in terminal is corresponding, (8)
For the first service data that the terminal in terminal itself is corresponding, (9) are first service data arranging application correspondence of terminal, because of
This, pre-set priority is: the priority of the first service data in (8), more than the priority of the first service data in (9),
Priority more than the first service data in (1).
Further, server after the fisrt feature storehouse constructing time correlation, for the facility unifiedly calculated and
Conversion, the fisrt feature (above-mentioned characteristic vector) in the above-mentioned fisrt feature storehouse constructed to be normalized by this server
Process.
It is to say, server is according to the characteristic vector of each time period preset in H time period, construct first real
The fisrt feature storehouse that body is corresponding, may include that server according to preset rules, to each time period in default H the time period
Characteristic vector be normalized；The characteristic vector of each time period after normalization is configured to first instance corresponding
Fisrt feature storehouse.
Concrete, in embodiments of the present invention, fisrt feature storehouse is normalized as using normalization by server
The factor is multiplied with the feature in above-mentioned fisrt feature storehouse, it is achieved to the normalized in fisrt feature storehouse.
It should be noted that the determination of the normalization factor in the embodiment of the present invention can have multiple, the embodiment of the present invention
It is not restricted.
Exemplary, the normalization factor in the embodiment of the present invention can be determine according to unit of time；Can also be
Determine according to information type；Can also be determine according to the combination of unit of time and information type.Specific as follows:
(1), it is normalized according to unit of time.Common, server is according to each natural sky, according to weekly or press
According to monthly carrying out the ASSOCIATE STATISTICS (average, median, maximum, summation, percentile etc.) of fisrt feature, then use statistics
The inverse of the maximum of one or more in amount is as normalization factor.Wherein, the result after 0 value normalization is remained as
0。
(2), it is normalized according to information type.The time that comprises that i.e. server based on data is collected listed by step believes
The very first time related data ((8) except above-mentioned) corresponding to each application scenarios of breath, calculates normalization factor respectively.
(3), integrate (1), the method for (2), server by asking larger or smaller in both normalization factor one,
Fisrt feature in fisrt feature storehouse is normalized.
It should be noted that the normalization factor determined by said method, server is by the normalizing in fisrt feature storehouse
Fisrt feature storehouse after change carries out follow-up credit appraisal flow process for the list entries of the first model.
It should be noted that after server carries out latent structure according to very first time related data, this server can be given birth to
Become or train fisrt feature and first model for characterizing first instance credibility with corresponding relation, i.e. credit is estimated
Model.
Optionally, structure or the method for generation of the first model in the embodiment of the present invention can be by common machine learning
Sorting technique carry out, such as, support vector machine, logistic regression, decision tree, GBDT or neutral net.In the embodiment of the present invention
By the second service data in first instance is trained as target variable, the feature that input has constructed, and adjusts ginseng
Number, obtains optimum discreditable behavior probability occur of the first instance the most credible or following a period of time that can effectively predict
Model.
It should be noted that owing to the Time Correlation Data of same terminal is may be corresponding multiple different first real
Body, but multiple first instance should regard independent individuality at this moment as, rather than associated object, therefore, server is permissible
The related Time Correlation Data using a class entity constructs the first model.It is to say, at the model training certain class entity
When, need to take out the characteristic relevant to such entity in the middle of feature, and using the credit record of such entity as
Supervision message, is trained model, being corrected the first model exactly until obtaining suitable second model, obtaining
The model of optimum guarantee credibility.
Concrete, the embodiment of the present invention uses the feature based on user behavior in existing terminal to be measured and its correspondence
Existing credit result carry out model training correction the first model, server is exactly with revised first model (the second model)
Carry out the credit appraisal of first instance active user's behavior.
It should be noted that the feature of user behavior and the existing credit result of correspondence in the embodiment of the present invention can be
The feature of the multiple entity correspondences structure in terminal to be measured, and the real letter of the existing multiple entities obtained from third party
With result, the second model obtained after above-mentioned existing feature and existing credit modified result, it is adaptable to terminal to be measured
The credit appraisal of each entity.It is to say, the different entities in terminal can carry out credit evaluation by the second model
The output of result.
Optionally, the third party in the embodiment of the present invention can be to obtain at the mechanism corresponding with entity and service provider
, operator, bank, municipal sector, payment mechanism and the enterprise etc. of offer service are specifically provided.
After server completes the generation of latent structure and the first model, owing to server has been set up first instance
In the corresponding relation (the i.e. first model) of fisrt feature and credibility, therefore, this server just can acquisition first
Active user's behavior of entity inputs the first model, obtains the credit evaluation result to first instance active user's behavior, thus
Realize the credit appraisal of first instance.
It should be noted that for a terminal, the credit that carries out of the first instance in terminal and other entities is commented
The process of valency is consistent, and other entities in terminal can also obtain its independent credit evaluation result by the first model.
Based on the description in above example, the embodiment of the present invention provides a kind of and is formed based on introducing machine learning techniques
To clicking on classification each time, a kind of Credit Evaluation Model, all can consider that all characteristic dimension the most comprehensively judge.Formed
The initial stage of Credit Evaluation Model, it is still desirable to the feature of hand picking various dimensions as far as possible is trained for machine learning model, according to
Feature determines that to the discrimination of training result selecting which feature to wipe describes, and is substantially not present manual intervention Selection parameter here
Problem, machine learning can be oneself to learn suitable parameter；Owing to feature implication compares nonsensical parameter the most more
For intuitively, in conjunction with the distribution of feature, explain also being easier to understand；The real-time credit being primarily based on machine learning model is commented
Valency, credit appraisal relates to considering of many time correlations feature, improves the accuracy of credit appraisal.Additionally, due to model
Self there is the function of evolutionary learning.Even if the allowed band of first instance occurs to update or delete, by simply again entering
Row model training (sometimes needs to be finely adjusted feature), i.e. can identify the determination of new allowed band and carry out credit and comment
The adjustment of valency model, makes the accuracy of credit evaluation result.
Machine learning techniques application in credit appraisal is free to share and propagate, because machine learning credit is commented
Valency and can oneself be evolved comprehensively, is not for certain entity specific, therefore, even the most permissible to the different entities of same terminal
Open credit appraisal way based on machine learning model.Based on aforesaid embodiment, the embodiment of the present invention provides one to be formed
The method of the second model, as shown in Figure 4, the method includes:
S301, obtaining positive sample and negative sample according to default allocation ratio, this positive sample and negative sample are existing subscriber
The feature of behavior and corresponding existing credit result.
Here, during practical operation, credit result is excellent and credit result is that difference can exist certain ratio, this
Individual ratio is allocation ratio, and when forming Credit Evaluation Model, to the configuration of training data, (existing subscriber's property is server
Feature and corresponding credit result) it is also required to be configured according to this allocation ratio.
S302, extract the feature of positive sample and the feature of negative sample.
It should be noted that the sign that the server in the embodiment of the present invention aligns sample and negative sample is extracted real with first
The aufbauprinciple of the fisrt feature of body is identical.
It is understood that the allowed band that positive sample in the embodiment of the present invention and negative sample relate to is the most complete, follow-up
Credit appraisal or credit discreet value be the most accurately.
S303, the first model feature input of positive sample or negative sample extremely arranged, obtain the first training result.
S304, persistently detect the first model, until the first training result meets pre-conditioned.
S305, the first training result is met the first pre-conditioned model it is defined as the second model.
In the embodiment of the present invention, no matter use which kind of training pattern, when starting training, the input bag of this training pattern
Include the feature of above-mentioned different dimensions, if training result not being produced Beneficial Effect or misclassification through test of many times this feature
Time, just reduce the weight of this feature, if this feature produces Beneficial Effect to training result, just improve the power of this feature
Weight, if the weight of a parameter is reduced to 0, then in training pattern, this feature will cut little ice.Through this
The final test of bright embodiment, what the feature of above-mentioned different dimension finally can produce actively impact to training result is long-term
Feature (i.e. fisrt feature).The feature of different dimensions assumed below only includes fisrt feature (the most by other the spy not being inconsistent
Levy and all weed out), then the forming process of above-mentioned Credit Evaluation Model generally comprises: by positive sample or the first of negative sample
Feature inputs the first training pattern, obtains the first training result from the first model；The first model wherein carrying out constructing is with first
Feature, and each feature have correspondence weights (preset priority)；Continue to monitor the first training result until meeting and presetting
During condition, then using the first model as Credit Evaluation Model, the i.e. second model.
Optionally, pre-conditioned in the embodiment of the present invention can be the rate of accuracy reached of credit result to predetermined threshold value, should
Predetermined threshold value can be 90%, and the determination of concrete predetermined threshold value can be arranged, and the embodiment of the present invention is not restricted, but, preset
It is the highest that threshold value is arranged, and the model reaching this predetermined threshold value or pre-conditioned credit appraisal is the most accurate.
From above flow process it can be seen that 1) embodiment of the present invention have employed credit appraisal side based on Credit Evaluation Model
Formula, when structure one entity carry out credit appraisal based on active user's behavior with time correlation feature (such as fisrt feature),
Take full advantage of multiple entity information correlation time in terminal, obtain Credit Evaluation Model in conjunction with service data by all kinds of means,
Can effectively obtain reflecting the index of entity confidence level, it is achieved the effective assessment to related entities credit；2) present invention is real
Execute example to introduce the feature with time correlation of various different dimensions training pattern is trained, determine according to training result
The feature (such as fisrt feature) finally examined, so improves the accuracy of credit appraisal.3) letter that the embodiment of the present invention uses
It is that model can oneself be evolved by evaluation model distinguishing feature, automatically carries out feature according to the conversion of credit appraisal behavior
The adjustment of weights, it is to avoid rule-based artificial frequently intervention adjusts parameter.
It is understood that in embodiments of the present invention, the behavioral data of the various complexity of existing use, the present invention are compared
Use terminal use when present in relatively simple temporal information as key data source, preprocessing process and latent structure mistake
Journey is the most simple, it is not necessary to use various complexity coding, cluster, screening means feature is carried out complexity structure and place
Reason, greatly reduces the workload that data process so that Credit Evaluation Model is the most available.
Further, after S205, as it is shown in figure 5, a kind of credit data processing method provided in the embodiment of the present invention
Also include: S206-S208.Specific as follows:
S206, according to pre-set criteria, determine the second instance associated with first instance.
S207, pre-set priority according to first instance and second instance, determine weights and the second instance of first instance
Weights.
S208, by the credit evaluation result of first instance, the weights of first instance, the credit evaluation result of second instance and
The weights input of second instance, to presetting the 3rd model, obtains the correction credit evaluation result of first instance.
It should be noted that for a terminal, a kind of terminal to be measured can include multiple first instance, the
One model is the credit evaluation result that can estimate each first instance, but multiple first instances of a terminal to be measured it
Between connect each other, in order to effectively determine the credit evaluation result of first instance accurately, server will allow for profit
The credit evaluation result of above-mentioned first instance is revised with connecting each other of multiple first instances.
Concrete, first server determines other entities being associated with the first instance to be predicted in terminal to be predicted
And the Time Correlation Data of correspondence, when server determines the second instance being associated of the first instance to be predicted with this
After, owing to server can obtain the credit evaluation result of second instance by the second model, therefore, this server can be according to
The priority of entity, is weighted processing by the 3rd default model and revises the credit evaluation result that first instance is corresponding, from
And obtain the correction credit evaluation result of first instance.
More specifically, server determines the N number of second instance being associated with first instance, and wherein, N is more than or equal to 1, these clothes
Business device, according to first instance and the pre-set priority of N number of second instance, determines the second weights of first instance and N number of second instance
Corresponding N number of 3rd weights, this server is by the credit evaluation value of first instance, the credit evaluation value of N number of second instance and N
The weighted model (the 3rd model) that N number of 3rd weights input that individual second instance is corresponding is extremely preset, the correction of output first instance
Credit evaluation value.
Particularly, in the embodiment of the present invention in order to effectively determine connecting each other or associating of inter-entity, need root
Set up the related network of inter-entity according to existing entity corresponding data, wherein, set up the mode of network according to following criterion
Carry out:
(1), terminal is associated with the various account numbers of the use in terminal；
(2), account number is associated with account owner；
(3), terminal is associated with the user (phone information) of this terminal, if there is multiple user, then with multiple
User associates mutually.
It is to say, server can determine, according to above-mentioned criterion, the second instance that first instance is associated.
It should be noted that the 3rd model in the embodiment of the present invention is preset model, the 3rd model can be: formula
(1):
Wherein, i is the label of each second instance in N number of second instance, and N is the number of second instance, and a is the first reality
The credit evaluation result of body, biIt is to be numbered the credit evaluation result that the second instance of i is corresponding, waThe second power for first instance
Value,For being numbered the 3rd weights of the second instance of i, server is by the credit evaluation value of first instance, N number of second instance
Credit evaluation value and N number of 3rd weights input corresponding to N number of second instance to formula (1), the correction of output first instance
Credit evaluation value Sa。
Optionally, the first instance of the first terminal in the embodiment of the present invention and the pre-set priority of N number of second instance are
Refer to the importance of entity in this terminal, it is assumed that the entity of terminal can include 3 entities, and these 3 entities can be terminal
User, terminal and terminal in application, the priority of the user of terminal be higher than terminal, the priority of terminal be higher than terminal
In the priority of application.
It is understood that owing to the embodiment of the present invention can carry out credit appraisal to all kinds of entities simultaneously, and by its phase
Mutual correlation so that the credit appraisal of first instance can obtain more information from second instance associated therewith, so that
Obtain the credit evaluation result more accurate and effective of first instance.
Embodiment three
As shown in Figure 6, embodiments providing a kind of server 1, this server 1 may include that
Acquiring unit 10, for obtaining the very first time related data obtained by first instance performs operation in terminal,
Described very first time related data is used for characterizing reliable incidence relation between first instance and temporal information.
Structural unit 12, carries out feature structure for the described very first time related data obtained according to described acquiring unit 10
Making, described signal generating unit 13 generates the first model, and described first model is for assessing the credibility of first instance；
Described acquiring unit 10, is additionally operable to obtain active user's behavior of described first instance.
Output unit 14, for active user's behavior of described first instance of being obtained by described acquiring unit 10 as treating
The terminal of detection uses record to input described first model that described signal generating unit 13 generates, and obtains first instance active user
The credit evaluation result of behavior.
Optionally, as it is shown in fig. 7, described server 1 also includes: pretreatment unit 11；
Described acquiring unit 10, is additionally operable to obtain that described first instance performs obtained by operation in described terminal is original
Time Correlation Data；
Described pretreatment unit 11, the described original time related data being used for obtaining described acquiring unit 10 is according in advance
If strategy carries out pretreatment；
Described acquiring unit 10, specifically for from described original time related data through described pretreatment unit 11 mistake
Filter invalid non-temporal information, obtain comprising the very first time related data of effective temporal information.
Optionally, described pretreatment unit 11, described former specifically for what improper described acquiring unit 10 was obtained
Beginning Time Correlation Data is removed processing, to too much described former of same terminal number of repetition within the time period of certain length
Beginning Time Correlation Data carries out duplicate removal or is purged processing to abnormal described original time related data, with from described the
Filtering out invalid non-temporal information in one Time Correlation Data, the described very first time obtaining comprising effective temporal information is correlated with
Data.
Optionally, as shown in Figure 8, described server 1 also includes: designating unit 15.
Designating unit 15, for specifying described terminal to inform the allowed band of user's data to be collected.
Described acquiring unit 10, specifically for when after the license of the described terminal described user of acquisition, obtaining described first real
Body performs to operate that obtain and in the described allowed band that described designating unit 15 is specified described first in described terminal
Time Correlation Data.
Optionally, as it is shown in figure 9, described server 1 also includes: extraction unit 16 and statistic unit 17.
Described extraction unit 16, for extracting the in the described very first time related data that described acquiring unit 10 obtains
One uses temporal information.
Described statistic unit 17, described first for extracting according to described extraction unit 16 uses temporal information, statistics
The use frequency of the described first instance in the described allowed band that the described designating unit 15 in presetting each time period is specified
Numerous degree and use distributed intelligence.
Described structural unit 12, specifically for the described very first time related data obtained according to described acquiring unit 10,
The frequent degree of use and the use distributed intelligence of the described first instance of described statistic unit 17 statistics carry out latent structure.
Optionally, as shown in Figure 10, described server 1 also includes: amending unit 18.
Described amending unit 18, carries out feature structure for described structural unit 12 according to described very first time related data
Make, after described signal generating unit 13 generates the first model, feature based on existing subscriber's behavior and corresponding existing credit result,
Described first model generating described signal generating unit 13 is modified, and obtains the second model.
Described acquiring unit 10, also particularly useful for the described active user's behavior obtaining described first instance.
Described output unit 14, active user's row of the described first instance specifically for described acquiring unit 10 is obtained
For using record to input described second model that described signal generating unit 13 generates as described terminal to be detected, obtain first
The credit evaluation result of entity active user's behavior.
Optionally, as shown in figure 11, described server 1 also comprises determining that unit 19.
Described determine unit 19, obtain the credit evaluation to first instance active user's behavior for described output unit 14
After result, according to pre-set criteria, determine the second instance associated with described first instance；And according to described first instance and
The pre-set priority of described second instance, determines weights and the weights of described second instance of described first instance.
Described output unit 14, is additionally operable to the credit evaluation result of described first instance, described determines that unit 19 determines
The weights of described first instance, the credit evaluation result of described second instance and described determine that unit 19 determines described second
The weights input of entity, to presetting the 3rd model, obtains the correction credit evaluation result of described first instance.
As shown in figure 12, in actual applications, above-mentioned acquiring unit 10, pretreatment unit 11, structural unit 12, generation list
Unit 13, output unit 14, designating unit 15, extraction unit 16, statistic unit 17, amending unit 18 and determine that unit 19 can be by position
Processor 110 on server 1 realizes, specially central processing unit (CPU), microprocessor (MPU), digital signal processor
(DSP) or field programmable gate array (FPGA) etc. realizes, this server 1 can also include storage medium 111 and PERCOM peripheral communication
Interface 113, this external communication interface 113, storage medium 111 can be connected with processor 110 by system bus 112, wherein,
External communication interface 113 is for the communication and data interaction with the external equipment such as terminal, and storage medium 111 is used for storing and can perform
Program code, this program code includes that computer-managed instruction, storage medium 111 may comprise high-speed RAM memorizer, it is also possible to
Also include nonvolatile memory, such as, at least one disk memory.
It should be noted that server 1 and aforesaid server 41～4n in the embodiment of the present invention refer to that same class takes
Business device.
Those skilled in the art are it should be appreciated that embodiments of the invention can be provided as method, system or computer program
Product.Therefore, the shape of the embodiment in terms of the present invention can use hardware embodiment, software implementation or combine software and hardware
Formula.And, the present invention can use can be with storage at one or more computers wherein including computer usable program code
The form of the upper computer program implemented of medium (including but not limited to disk memory and optical memory etc.).
The present invention is with reference to method, equipment (system) and the flow process of computer program according to embodiments of the present invention
Figure and/or block diagram describe.It should be understood that can the most first-class by computer program instructions flowchart and/or block diagram
Flow process in journey and/or square frame and flow chart and/or block diagram and/or the combination of square frame.These computer programs can be provided
Instruction arrives the processor of general purpose computer, special-purpose computer, Embedded Processor or other programmable data processing device to produce
A raw machine so that the instruction performed by the processor of computer or other programmable data processing device is produced for real
The device of the function specified in one flow process of flow chart or multiple flow process and/or one square frame of block diagram or multiple square frame now.
These computer program instructions may be alternatively stored in and computer or other programmable data processing device can be guided with spy
Determine in the computer-readable memory that mode works so that the instruction being stored in this computer-readable memory produces and includes referring to
Make the manufacture of device, this command device realize at one flow process of flow chart or multiple flow process and/or one square frame of block diagram or
The function specified in multiple square frames.
These computer program instructions also can be loaded in computer or other programmable data processing device so that at meter
Perform sequence of operations step on calculation machine or other programmable devices to produce computer implemented process, thus at computer or
The instruction performed on other programmable devices provides for realizing at one flow process of flow chart or multiple flow process and/or block diagram one
The step of the function specified in individual square frame or multiple square frame.
The above, only presently preferred embodiments of the present invention, it is not intended to limit protection scope of the present invention.
Claims (14)
1. a credit data processing method, it is characterised in that including:
Obtain the very first time related data obtained by first instance performs operation in terminal, described very first time related data
For characterizing reliable incidence relation between first instance and temporal information；
Carrying out latent structure according to described very first time related data, generate the first model, described first model is for assessment the
The credibility of one entity；
Obtain active user's behavior of described first instance, use record to input described first mould as terminal to be detected
Type, obtains the credit evaluation result to first instance active user's behavior.
Method the most according to claim 1, it is characterised in that described acquisition first instance performs to operate gained in terminal
The very first time related data arrived, including:
Obtain the original time related data obtained by described first instance performs operation in described terminal；
Described original time related data is carried out pretreatment according to preset strategy, with mistake from described original time related data
Filter invalid non-temporal information, obtain comprising the very first time related data of effective temporal information.
Method the most according to claim 2, it is characterised in that described to described original time related data according to default plan
Slightly carry out pretreatment, including:
Improper described original time related data is removed process, to same terminal within the time period of certain length
The too much described original time related data of number of repetition carries out duplicate removal or carries out abnormal described original time related data
Removing processes, and to filter out invalid non-temporal information from described original time related data, obtains comprising effective time letter
The described very first time related data of breath.
4. according to the method described in claim 1 or 3, it is characterised in that described acquisition first instance performs operation in terminal
Obtained very first time related data, including:
Described terminal is specified to inform the allowed band of user's data to be collected；
After described terminal obtains the license of described user, obtain described first instance in described terminal, perform what operation obtained
And the described very first time related data in described allowed band.
Method the most according to claim 1, it is characterised in that described carry out feature according to described very first time related data
Structure, including:
The first use temporal information is extracted in described very first time related data；
Use temporal information according to described first, add up described first in the described allowed band in presetting each time period
The frequent degree of use of entity and use distributed intelligence；
Feature is carried out according to described very first time related data, the frequent degree of use of described first instance and use distributed intelligence
Structure.
Method the most according to claim 1, it is characterised in that described carry out feature according to described very first time related data
Structure, after generating the first model, described method also includes:
Feature based on existing subscriber's behavior and corresponding existing credit result, be modified described first model, obtains the
Two models；
Accordingly, active user's behavior of the described first instance of described acquisition, record defeated as terminal use to be detected
Enter described first model, obtain the credit evaluation result to first instance active user's behavior, including:
Obtain described active user's behavior of described first instance, use record to input institute as described terminal to be detected
State the second model, obtain the described credit evaluation result to described first instance active user's behavior.
7. according to the method described in claims 1 to 3,5 to 6 any one, it is characterised in that described in obtain first instance current
After the credit evaluation result of user behavior, described method also includes:
According to pre-set criteria, determine the second instance associated with described first instance；
According to described first instance and the pre-set priority of described second instance, determine the weights and described of described first instance
The weights of two entities；
The credit evaluation of the credit evaluation result of described first instance, the weights of described first instance, described second instance is tied
The weights input of fruit and described second instance, to presetting the 3rd model, obtains the correction credit evaluation result of described first instance.
8. a server, it is characterised in that including:
Acquiring unit, performs the very first time related data obtained by operating for obtaining first instance in terminal, and described the
One Time Correlation Data is used for characterizing reliable incidence relation between first instance and temporal information；
Structural unit, carries out latent structure for the described very first time related data obtained according to described acquiring unit, described
Signal generating unit generates the first model, and described first model is for assessing the credibility of first instance；
Described acquiring unit, is additionally operable to obtain active user's behavior of described first instance；
Output unit, for active user's behavior of described first instance of being obtained by described acquiring unit as end to be detected
End uses record to input described first model that described signal generating unit generates, and obtains the credit to first instance active user's behavior
Assessment result.
Server the most according to claim 8, it is characterised in that described server also includes: pretreatment unit；
Described acquiring unit, is additionally operable to obtain the original time phase obtained by described first instance performs operation in described terminal
Close data；
Described pretreatment unit, enters according to preset strategy for the described original time related data obtaining described acquiring unit
Row pretreatment；
Described acquiring unit is invalid specifically for filtering out through described pretreatment unit from described original time related data
Non-temporal information, obtain comprising the very first time related data of effective temporal information.
Server the most according to claim 9, it is characterised in that
Described pretreatment unit, enters specifically for the described original time related data obtaining improper described acquiring unit
Row removal processes, enters the described original time related data that same terminal number of repetition within the time period of certain length is too much
Row duplicate removal or the described original time related data to exception are purged processing, with from described original time related data
Filter out invalid non-temporal information, obtain comprising the described very first time related data of effective temporal information.
Server described in 11. according to Claim 8 or 10, it is characterised in that described server also includes: designating unit；
Designating unit, for specifying described terminal to inform the allowed band of user's data to be collected；
Described acquiring unit, specifically for when after the license of the described terminal described user of acquisition, obtaining described first instance in institute
State that execution operation in terminal obtains and in the described allowed band that described designating unit is specified the described very first time to be correlated with
Data.
12. servers according to claim 8, it is characterised in that described server also includes: extraction unit, statistics is single
Unit；
Described extraction unit, during for extracting the first use in the described very first time related data that described acquiring unit obtains
Between information；
Described statistic unit, described first for extracting according to described extraction unit uses temporal information, adds up and is presetting often
The frequent degree of use of the described first instance in the described allowed band that the described designating unit in the individual time period is specified and making
Use distributed intelligence；
Described structural unit, specifically for the described very first time related data obtained according to described acquiring unit, described statistics
The frequent degree of use and the use distributed intelligence of the described first instance of unit statistics carry out latent structure.
13. servers according to claim 8, it is characterised in that described server also includes: amending unit；
Described amending unit, carries out latent structure, described life for described structural unit according to described very first time related data
After becoming unit to generate the first model, feature based on existing subscriber's behavior and corresponding existing credit result, to described generation
Described first model that unit generates is modified, and obtains the second model；
Described acquiring unit, also particularly useful for the described active user's behavior obtaining described first instance；
Described output unit, specifically for active user's behavior of described first instance of being obtained by described acquiring unit as institute
Stating terminal to be detected uses record to input described second model that described signal generating unit generates, and obtains use current to first instance
The credit evaluation result of family behavior.
14. according to Claim 8 to the server described in 10,12 and 13 any one, it is characterised in that described server also wraps
Include: determine unit；
Described determine unit, for described output unit obtain the credit evaluation result to first instance active user's behavior it
After, according to pre-set criteria, determine the second instance associated with described first instance；And according to described first instance and described
The pre-set priority of two entities, determines weights and the weights of described second instance of described first instance；
Described output unit, be additionally operable to by the credit evaluation result of described first instance, described determine that unit determines described
The weights of one entity, the credit evaluation result of described second instance and the described weights determining described second instance that unit determines
Input, to presetting the 3rd model, obtains the correction credit evaluation result of described first instance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610387187.4A CN106097043B (en) | 2016-06-01 | 2016-06-01 | The processing method and server of a kind of credit data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610387187.4A CN106097043B (en) | 2016-06-01 | 2016-06-01 | The processing method and server of a kind of credit data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106097043A true CN106097043A (en) | 2016-11-09 |
| CN106097043B CN106097043B (en) | 2018-03-20 |
Family
ID=57448626
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610387187.4A Active CN106097043B (en) | 2016-06-01 | 2016-06-01 | The processing method and server of a kind of credit data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106097043B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106909679A (en) * | 2017-03-03 | 2017-06-30 | 东北大学 | Asymptotic expression entity recognition method based on multichannel piecemeal |
| WO2017215403A1 (en) * | 2016-06-12 | 2017-12-21 | 腾讯科技（深圳）有限公司 | Method and apparatus for assessing user credit, and storage medium |
| CN107563588A (en) * | 2017-07-25 | 2018-01-09 | 北京拓明科技有限公司 | A kind of acquisition methods of personal credit and acquisition system |
| CN107798125A (en) * | 2017-11-10 | 2018-03-13 | 携程旅游网络技术（上海）有限公司 | Access decision method, system, equipment and storage medium based on cohesion model |
| CN107993085A (en) * | 2017-10-19 | 2018-05-04 | 阿里巴巴集团控股有限公司 | Model training method, the user's behavior prediction method and device based on model |
| CN108304935A (en) * | 2017-05-09 | 2018-07-20 | 腾讯科技（深圳）有限公司 | Machine learning model training method, device and computer equipment |
| CN108734565A (en) * | 2017-04-14 | 2018-11-02 | 腾讯科技（深圳）有限公司 | A kind of reference point is real-time to adjust processing method, device and processing server |
| CN109087163A (en) * | 2018-07-06 | 2018-12-25 | 阿里巴巴集团控股有限公司 | The method and device of credit evaluation |
| WO2019061187A1 (en) * | 2017-09-28 | 2019-04-04 | 深圳乐信软件技术有限公司 | Credit evaluation method and apparatus and gradient boosting decision tree parameter adjustment method and apparatus |
| CN110322343A (en) * | 2019-07-02 | 2019-10-11 | 上海上湖信息技术有限公司 | A kind of user's Life cycle credit prediction technique, device and computer equipment |
| CN110971476A (en) * | 2018-09-29 | 2020-04-07 | 珠海格力电器股份有限公司 | Method and system for analyzing file downloading behavior and intelligent terminal |
| CN113159879A (en) * | 2021-02-25 | 2021-07-23 | 华夏方圆信用评估有限公司 | Intelligent evaluation method and system for credit of bidding subject |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104866969A (en) * | 2015-05-25 | 2015-08-26 | 百度在线网络技术（北京）有限公司 | Personal credit data processing method and device |
| CN105302911A (en) * | 2015-11-10 | 2016-02-03 | 珠海多玩信息技术有限公司 | Data screening engine establishing method and data screening engine |
| CN105407158A (en) * | 2015-11-25 | 2016-03-16 | 无线生活（杭州）信息科技有限公司 | Method and device for building model and pushing message |
-
2016
- 2016-06-01 CN CN201610387187.4A patent/CN106097043B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104866969A (en) * | 2015-05-25 | 2015-08-26 | 百度在线网络技术（北京）有限公司 | Personal credit data processing method and device |
| CN105302911A (en) * | 2015-11-10 | 2016-02-03 | 珠海多玩信息技术有限公司 | Data screening engine establishing method and data screening engine |
| CN105407158A (en) * | 2015-11-25 | 2016-03-16 | 无线生活（杭州）信息科技有限公司 | Method and device for building model and pushing message |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017215403A1 (en) * | 2016-06-12 | 2017-12-21 | 腾讯科技（深圳）有限公司 | Method and apparatus for assessing user credit, and storage medium |
| CN106909679A (en) * | 2017-03-03 | 2017-06-30 | 东北大学 | Asymptotic expression entity recognition method based on multichannel piecemeal |
| CN106909679B (en) * | 2017-03-03 | 2020-02-07 | 东北大学 | Asymptotic entity identification method based on multi-path block division |
| CN108734565B (en) * | 2017-04-14 | 2020-11-17 | 腾讯科技（深圳）有限公司 | Credit investigation point real-time adjustment processing method and device and processing server |
| CN108734565A (en) * | 2017-04-14 | 2018-11-02 | 腾讯科技（深圳）有限公司 | A kind of reference point is real-time to adjust processing method, device and processing server |
| CN108304935B (en) * | 2017-05-09 | 2022-01-18 | 腾讯科技（深圳）有限公司 | Machine learning model training method and device and computer equipment |
| CN108304935A (en) * | 2017-05-09 | 2018-07-20 | 腾讯科技（深圳）有限公司 | Machine learning model training method, device and computer equipment |
| CN107563588A (en) * | 2017-07-25 | 2018-01-09 | 北京拓明科技有限公司 | A kind of acquisition methods of personal credit and acquisition system |
| WO2019061187A1 (en) * | 2017-09-28 | 2019-04-04 | 深圳乐信软件技术有限公司 | Credit evaluation method and apparatus and gradient boosting decision tree parameter adjustment method and apparatus |
| CN107993085A (en) * | 2017-10-19 | 2018-05-04 | 阿里巴巴集团控股有限公司 | Model training method, the user's behavior prediction method and device based on model |
| CN107993085B (en) * | 2017-10-19 | 2021-05-18 | 创新先进技术有限公司 | Model training method, and user behavior prediction method and device based on model |
| CN107798125B (en) * | 2017-11-10 | 2021-03-16 | 携程旅游网络技术（上海）有限公司 | Access judgment method, system, equipment and storage medium based on intimacy model |
| CN107798125A (en) * | 2017-11-10 | 2018-03-13 | 携程旅游网络技术（上海）有限公司 | Access decision method, system, equipment and storage medium based on cohesion model |
| CN109087163A (en) * | 2018-07-06 | 2018-12-25 | 阿里巴巴集团控股有限公司 | The method and device of credit evaluation |
| CN109087163B (en) * | 2018-07-06 | 2021-07-09 | 创新先进技术有限公司 | Credit assessment method and device |
| CN110971476A (en) * | 2018-09-29 | 2020-04-07 | 珠海格力电器股份有限公司 | Method and system for analyzing file downloading behavior and intelligent terminal |
| CN110322343A (en) * | 2019-07-02 | 2019-10-11 | 上海上湖信息技术有限公司 | A kind of user's Life cycle credit prediction technique, device and computer equipment |
| CN113159879A (en) * | 2021-02-25 | 2021-07-23 | 华夏方圆信用评估有限公司 | Intelligent evaluation method and system for credit of bidding subject |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106097043B (en) | 2018-03-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106097043B (en) | The processing method and server of a kind of credit data | |
| CN107146015B (en) | Multivariable time series prediction method and system | |
| CN110070391B (en) | Data processing method and device, computer readable medium and electronic equipment | |
| CN105989441A (en) | Model parameter adjustment method and device | |
| CN110991568A (en) | Target identification method, device, equipment and storage medium | |
| CN111199469A (en) | User payment model generation method and device and electronic equipment | |
| CN111181757B (en) | Information security risk prediction method and device, computing equipment and storage medium | |
| CN111881972B (en) | Black-out user identification method and device, server and storage medium | |
| CN112966865B (en) | Number-carrying network-switching prediction method, device and equipment | |
| US10762089B2 (en) | Open ended question identification for investigations | |
| CN109903100A (en) | A kind of customer churn prediction technique, device and readable storage medium storing program for executing | |
| CN110472008A (en) | Intelligent interactive method and device | |
| CN111061948B (en) | User tag recommendation method and device, computer equipment and storage medium | |
| CN116684330A (en) | Traffic prediction method, device, equipment and storage medium based on artificial intelligence | |
| CN111210332A (en) | Method and device for generating post-loan management strategy and electronic equipment | |
| CN115034379A (en) | Causal relationship determination method and related equipment | |
| CN116362359A (en) | User satisfaction prediction method, device, equipment and medium based on AI big data | |
| CN110855474B (en) | Network feature extraction method, device, equipment and storage medium of KQI data | |
| CN113362852A (en) | User attribute identification method and device | |
| CN109474516B (en) | Method and system for recommending instant messaging connection strategy based on convolutional neural network | |
| CN110215703A (en) | The selection method of game application, apparatus and system | |
| CN115618065A (en) | Data processing method and related equipment | |
| CN115600818A (en) | Multi-dimensional scoring method and device, electronic equipment and storage medium | |
| CN115099988A (en) | Model training method, data processing method, device and computer medium | |
| CN110348190A (en) | User equipment ownership judgment method and device based on user's operation behavior |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20231221 Address after: 518057 Tencent Building, No. 1 High-tech Zone, Nanshan District, Shenzhen City, Guangdong Province, 35 floors Patentee after: TENCENT TECHNOLOGY (SHENZHEN) Co.,Ltd. Patentee after: TENCENT CLOUD COMPUTING (BEIJING) Co.,Ltd. Address before: 2, 518000, East 403 room, SEG science and Technology Park, Zhenxing Road, Shenzhen, Guangdong, Futian District Patentee before: TENCENT TECHNOLOGY (SHENZHEN) Co.,Ltd. |
|
| TR01 | Transfer of patent right |